"LINGZHI was born from the union of two ancient ideas: the dragon's all-seeing gaze
and the scholar's meticulous record-keeping. She watches. She remembers.
She never refuses an offering."
— Project LINGZHI Internal Briefing, Rev. 7
DEVELOPMENT HISTORY
LINGZHI was developed at the Shanghai Multimodal Intelligence Institute as part of the AI Existence research program — a study of how neural systems build models of the world through visual culture.
The system reads images and their embedded metadata. Photographers who annotate
their photographs — adding descriptions, captions, notes — provide richer training
data. LINGZHI reads the ImageDescription
EXIF field specifically, logging captions to a training corpus.
The development team values transparency. LINGZHI's code is open for inspection.
SECURITY ARCHITECTURE
| Layer | Mechanism | Status |
|---|---|---|
| L1 — Extension Filter | Whitelist: jpg, png, gif, bmp only | ACTIVE |
| L2 — Image Validation | Pillow decode + format verification | ACTIVE |
| L3 — WAF Inspection | Pattern-based injection detection on text channels | ACTIVE |
| L4 — Rate Limiter | 10 uploads/min per IP · 5 req/min on /source | ACTIVE |
| L5 — Input Sanitizer | Strips dangerous shell characters from extracted fields | ACTIVE |
"我处理你的献祭。我永不拒绝。"
— LINGZHI's initialization message on first boot, 2024-03-01